Jarrar.lecture notes.aai.2011s.ontology part4_methodologies
0 likes441 views
The document discusses stepwise methodologies for building ontologies. It outlines common steps such as identifying the purpose and scope, capturing concepts and relationships, coding the ontology formally, integrating existing ontologies, evaluation, and documentation. It emphasizes starting with a middle-out approach to capture definitions and discusses reaching consensus among those involved in building the ontology. Modularization of ontologies into reusable components is also presented as an important aspect of the methodology.
Jarrar.lecture notes.aai.2011s.ontology part4_methodologies
- 1. Lecture Notes University of Birzeit2nd Semester, 2010Advanced Artificial Intelligence (SCOM7341)OntologyPart 4 Stepwise MethodologiesDr. Mustafa Jarrarmjarrar@birzeit.eduwww.jarrar.infoUniversity of Birzeit
- 2. Reading Material0) Everything in these slides1) Jarrar, M.: Towards Methodological Principles for Ontology Engineering. PhD thesis, Vrije Universiteit Brussel (2005). See http://www.jarrar.infoOnly chapter 4 (Section 1-3)2) Mustafa Jarrar: Towards Effectiveness and Transparency in e-Business Transactions, An Ontology for Customer Complaint Management . http://www.jarrar.info/publications/mjarrar-CCFORM-chapter.v08.pdfThis is a case study
- 3. MethodologyLet’s discuss form where to start, if you want to build an ontology for:E-government E-BankingE-HealthBioinformatics Multilingual search engine… What are the phases of the ontology development life cycle? taking into account that Ontologies might be built collaboratively by many people.
- 4. Methodological QuestionsHow can tools and techniques best be applied?
- 5. Which languages and tools should be used in which circumstances, and in which order?
- 6. What about issues of quality control and resource management?
- 7. Many Methodologies exist ! But non is good! Because each project/application/domain is different, and the background of the people involved are also different.
- 8. We will overview some common steps, try to learn smartly, and follow these steps literally. You should have your won methodology for each ontology.Most Methodologies propose:1- Identify Purpose and Scope2- Building the Ontology2.1- ontology capture2.2 ontology coding3- Integrating existing ontologies4- Evaluation5- Documentation
- 9. 1- Purpose and ScopeThere is no correct ontology of a specific domain An ontology is an abstraction of a particular domain, and there are always alternatives.What is included in this abstraction should be smartly determined by:the use to which the ontology will be put, such as:– Interoperability between systems.– improve quality Search.– Communication between people and organizations (important).by future extensions that are already anticipated.
- 10. 1- Purpose and ScopeWhen you specify the purpose and scope, you should specify the following:
- 11. What is the domain that the ontology will cover? The notion of context, in the double articulation theory, can be part of the Purpose and Scope. For what we are going to use the ontology?
- 12. For what types of questions should the ontology provide answers?
- 13. Who will use and maintain the ontology? And how?Be carful with the ontology usability - reusability trade-off
- 14. 2- Building the Ontology2.1- Ontology Capture– Identify key concepts and relationships.– Produce clear text definitions for these concepts (glosses, etc.).– Identify terms that refer to these concepts– Reach Consensus (Consensus is an indication of correctness).2.2- Ontology Coding/Specification/Characterization – Explicit representation of the “conceptualization” in some formal language.
- 15. 2.1- Ontology Capture: Scoping• Brainstorming– Produce all potentially relevant terms and phrases.Nouns form the basis for concept names
- 16. Verbs (or verb phrases) form the basis for property and names.– People involved must have substantial domain expertise.• Can we automate some steps to:Extraction the list of Concepts, and/or instances.
- 17. Extract candidate relations, and/or subsumptions.
- 18. Generate glosses.• Grouping: Structure terms loosely into work areas/topics– Provisionally categorize them for inclusion or exclusion (purpose and scope)– Keep notes of these decisions.– Group similar terms and potential synonyms together.
- 19. 2.1- Ontology Capture: Produce DefinitionsDetermine suitable meta-ontology
- 20. especially: use words and modeling primitives in a consistent manner (e.g. Type, role, entity, instance, relationship...)
- 21. Work Areas: Start with the most basic/important
- 22. Define the most basic (i.e. important) terms first in each work area before moving to more abstract or more specific terms.
- 23. Semantic overlap with others must be right in the first place, otherwise lot of redundant re-working.
- 24. Terms: Produce definitions in a middle-out fashion
- 25. rather than top-down or bottom up. – will be discussed later.Define TaxonomyRelevant terms must be organized in a taxonomic hierarchy (i.e., subsumptions )
- 26. Opinions differ on whether it is more efficient to do this in a top-down or a bottom-up fashion.
- 27. Ensure that hierarchy is indeed a taxonomy:
- 28. If A subsumes B, then every instance of A must also be a subsume B (compatible with semantics of rdfs:subClassOf
- 29. Insuring the correctness of subsumptions needs philosophical thinking (the OntoClean Methodology will be used.). Define PropertiesThe semantics of subsumption demands that whenever A subsumes B, every property that holds for instances of B must also apply to instances of A (called inheritance).
- 30. It makes sense to attach properties to the highest class in the hierarchy to which they apply.Define Properties (2)While attaching properties to concepts, it makes sense to immediately provide statements about the domain and range of these properties.
- 31. There is a methodological tension here between generality and specificity:
- 32. Flexibility (inheritance to subsumptions)
- 33. Detection of inconsistencies and misconceptionsAdd Rules and RestrictionsCardinality Restrictions
- 34. Which properties should be unique, mandatory, disjunctions, restricted values…etc.
- 36. symmetry, transitivity, inverse properties, functional values Define Some Important InstancesSome important instances (might) be added to the ontology, of needed. Such as:
- 38. Person: Arafat
- 39. Capital: JerusalemAdvantages of the Middle-out Approach• A bottom-up approach results in a high degree of detail– increases overall effort– makes it difficult to spot commonality between related concepts.– increases risk of inconsistencies and re-work• Top-down allow better control of degree of detail– risk of arbitrary high-level categories– risk of limited stability• Middle-out strikes a balance in terms of the level of detail• The higher level categories naturally arise and are thus more likely to be stable.
- 40. Reaching Agreement: Some suggestionsOntologies are made to be agreed and shared, thus it is VERY important to make that people agree on them.
- 41. How to facilitate reach agreement?• Produce a natural language text definitions• Ensure consistency with terms already in use– use existing thesauri and dictionaries– avoid introducing new terms in the definitions• Indicate relationships with other commonly used terms– synonyms, variants, such referring to different dimensions• Give examples
- 42. Integrating Existing Ontologies• Check overlap with existing ontologies• Establish formal links– Produce mappings to existing concept definitions– Import and extend existing ontologies• Avoid re-inventing the wheel!
- 43. Ontology EvaluationSeveral Type of evaluations:Usability Evaluation: Validate whether the ontology produced is consistent with and meets the requirements specification.Syntax evaluation: Validate whether the ontology well-formed w.r.t the used language.Logical evaluation: Validate whether the ontology has axioms contradicting or implying each other.Ontological Evaluation: Validate whether the ontology has concepts that should be instances, sub-concepts that should be roles, etc. (The OntoClean methodology is very good for this evaluation)
- 44. Check for Implications and ContradictionsSome tools exist to automatically detect logical correctness (contradictions and implications), depending on the used ontology language (Such as ORM: DogmaModeler, OWL: Racer)
- 45. Some GuidelinesClarity: The ontology engineer should communicate effectively with the domain experts (= ask the rightquestions):– Natural language definitions.– Give examples, alternative, and contradictions, elicit knowledge– emphasize distinctionsCoherence: The ontology should be internally consistent Syntactically correct.Logical consistent.Ontologically consistent.Extensibility: modularize the ontology in a way it is easy to build, understand, and maintain. What should be in a module?Reusability and Usability: be innovative to tradeoff this smartly.
- 47. ModularizationDevelop an application axiomatization as a set of modules and later compose to form one module.Domain AxiomatizationApplication AxiomatizationModularizationWhy to modularize?Because Modules are:Easier to reuseEasier to build, maintain, and replaceEnable distributed development of modulesEnable the effective management and browsingWhen to modularize?Modularity criteria: Subject-oriented Purpose/Task-oriented Stability